• This paper proposes a correlation-adaptive polygonal affine optimization framework that fuses historical and forecast data to dynamically tighten uncertainty sets, eliminating low-probability extremes without sacrificing robustness. • This paper integrate adjustable electrolytic aluminum loads—modeled via coupled power-current-temperature constraints—and battery storage to enhance renewable accommodation and reduce operating costs. To address the conservatism in economic dispatch caused by source-load uncertainties and correlations among renewable energy sources in industrial park energy systems, this paper proposes a correlation-adaptive polygonal affine optimization dispatch method.First, mathematical models are developed for high-energy-intensity aluminum electrolysis loads, thermal power units, and energy storage systems.Second, based on affine arithmetic theory, an enhanced parallelogram correlation model is constructed to represent uncertainties in wind and photovoltaic power generation as well as load demand. Historical and forecast data are utilized to dynamically adjust the polygonal uncertainty sets, thereby excluding low-probability extreme scenarios.By integrating the synergistic interaction mechanism between aluminum electrolysis loads and energy storage systems, an optimization dispatch model is formulated with the objective of minimizing operational costs.Simulation results show that, compared with conventional interval affine optimization, the proposed method significantly reduces dispatch conservatism while maintaining production continuity and economic feasibility. It effectively improves the accommodation rate of renewable energy and narrows the uncertainty interval of operational costs.The proposed approach provides a high-precision, low-conservatism solution for multi-energy coordinated optimization in industrial parks. Moreover, as the output of day-ahead dispatch, it can serve as secondary boundaries for intraday rolling optimization, thereby supporting preparatory decision-making for short-term electricity trading.
Yang et al. (Fri,) studied this question.